ICRA 2017 Papers

In order to achieve harmony in the mixed traffic, it is crucial to have autonomous vehicles behave like human drivers. This work addresses a vision-based approach toward human-like lane following behavior in complex urban environment. At first, a deep architecture is adopted to generate a set of vehicle hypotheses. Subsequently, a hybrid merging procedure is performed to jointly output the final detection results based on both the image evidence and the statistical support of vehicle hypotheses. After that, the detected vehicles are classified into six categories by Bayesian Network, i.e., leader vehicle, parking vehicle, tail-end vehicle, exiting vehicle, merging vehicle and other vehicle. With this information, a learning-based instance-level behavior-induction potential map is constructed to generate a safe as well as reasonable local path for following a predefined lane-level route. Experimental results in various typical but challenging urban traffic scenes substantiated the effectiveness of the proposed approach.

In this paper we use KUKA's Fast Robot Interface (FRI) to design and implement a tracking controller on the Lightweight Robot (LBR) IIWA. We seek low latency, accurate and smooth tracking of the link positions to facilitate human interaction tasks. Focusing on a single joint and its low-level series elastic dynamics, we identify the internal torque control structure and its characteristics. Tracking controllers of varying complexity are tested in an optical motion capture system to provide an independent external reference measurement. Using full state feedback of both motor position and sensed joint torque, we achieve smooth and good tracking of the unsensed link positions.

We propose a novel hydraulic hybrid servo drive for robotic applications. This method embeds a small servo-controlled pump into a hydraulic metering circuit. The circuit is compactly integrated into a servo-unit, then replicated for each joint, and connected to one common low-pressure line. Thanks to the boosting effect, one can simultaneously achieve high-load and high-precision servo control performance with low-cost components. The paper describes the principle of the new circuit and two realizations: one for a slider testbed with a single-rod cylinder, the other for a three-joint manipulator prototype. The slider experiments show that the proposed circuit can not only achieve the high piston velocity, but also exclusively generate large piston force with a small positional error up to the resolution of the servo-pump. Moreover, experiments on the manipulator demonstrate that the robot can compliantly respond to external perturbations, thanks to the high backdrivability of the servo-pump.

Performing detailed work on objects requires precise localization. Currently humans aid machines in localization either by direct operation, or implicitly by designing a sequence of actions a robot follows. Our approach to automate localization is to reason over many potential actions, perform the best information gathering action, and then use the measurement obtained to update a non-Gaussian belief. We propose a method for autonomous localization of objects with initial 6DOF uncertainty capable of reasoning about and performing measurements with low uncertainty and arbitrary error models. Surprisingly, common methods capable of modeling arbitrary belief distributions perform poorly as measurement uncertainty decreases, so we modify a particle filter to handle these accurate measurements produced by tactile or laser sensors. We then show how the expected information gain of the proposed measurement can be calculated efficiently from these particles. We present experiments, both in simulation and on hardware, that show our method is both fast and accurate.

Constraint on the actuation and power resources is often the critical limiting factor for a robot to perform desired tasks. Increasing torque and energy capacity may be a solution, but is seldom viable for robots already built. An attractive alternative is to carefully generate motion trajectories that maximally leverages upon the limited torque and energy resources. In this endeavor, singularity, which is deemed undesirable due to lose of manipulability, could be utilized to an advantage. This paper presents analysis of force and momentum generated through contact in relation to the singularity. The analysis shows that a motion at or near singularity not only maximally leverages the torque limits to generate forces in quasi-static motions, but is also optimally energy efficient for dynamical motion when it comes to momentum generation. Based on a simplified model, we discuss mechanical advantage aspects of a robotic leg and describe range of feasible forces that can be generated together with directions in which singular position becomes minimum torque configuration. Then we define stroke motion and establish upper bounds on the momentum generated through contact. Collinear stroke, where motion is along a straight line, is examined with respect to singularity.

This paper, for the first time, describes the estimation of the time-varying impedance of the human ankle in the sagittal (SP) and frontal (FP) planes during the stance phase of walking. The result of this work is aimed to provide design parameters for the development of 2-DOF powered ankle-foot prostheses capable of mimicking the time-varying impedance of the human ankle. Sixteen axes of rotations combining different amounts of SP and FP rotations were studied. For each axis, positive and negative rotations were considered separately. Four unimpaired male subjects walked on an instrumented vibrating platform that applied combined torque perturbations in the SP and FP simultaneously, while the ankle angles and torques were recorded. Based on the recorded data, the ankle impedance was estimated with a time resolution of 20 ms from 7% to 93% of the stance length (SL). The ankle stiffness and damping showed great variability through the SL and across axes of rotation. The maximum stiffness was 4.7±0.5 Nm/rad/kg at 0.21 s of the SL when the ankle rotated at an axis 22.5° from the SP combining Dorsiflexion (D) and Inversion (I). The minimum stiffness was 1.4±0.6 Nm/rad/kg at 0.05 s of the SL at an axis 45° from the SP combining D and Eversion (E). The maximum damping was 0.09±0.02 Nms/rad/kg at 0.21 s of the SL combining D and I at an axis 25° from the SP. The minimum was 0.02±0.01 Nms/rad/kg at 0.05 s of the SL combining P and I at an axis 45° from the SP.

Effective in situ, in vivo tumour margin assessment is an important, yet unmet, clinical demand in surgical oncology. Recent advances in probe-based optical imaging tools such as confocal endomicroscopy is making inroads in clinical applications. In practice, maintaining consistent tissue contact whilst ensuring large area surveillance is crucial for its practical adoption and for this reason there is a great demand for robotic assistance so that high-speed endomicroscopes can be combined with autonomous scanning, thus simplifying its incorporation in routine surgical workflows. In this paper, a cooperatively controlled robotic manipulator is developed, which provides a stable mechatronically-enhanced platform for micro-scanning tools to perform local high resolution mosaics over 3D undulating moving surfaces. Detailed kinematic and overall system performance analyses are provided and the results demonstrate the adaptability in terms of both contact force and orientation control of the system, and thus its simplicity in practical deployment and value for clinical adoption.

Magnetic micromanipulation has the advantage of untethered control, high precision, and biocompatibility and has recently undergone great advances. The magnetic micromanipulation task to tackle in this work is to three-dimensionally navigate a 5-micrometer magnetic bead inside a mouse embryo and perform mechanical measurements at multiple locations. Existing technologies are not able to achieve these navigation and measurement goals because of poor magnetic force scaling and/or lacking the capability of applying an accurately controlled force. This paper reports a robotic magnetic tweezer system that enables, for the first time, intra- embryonic magnetic navigation and force application. A single magnetic bead was introduced into a mouse embryo via robotic microinjection. The robotic magnetic tweezer system accurately controls the position of the magnetic bead via visually servoed magnetic control. The system is also capable of applying forces up to 120 pN with a resolution of 1.78 pN for performing mechanical measurements on the cellular structures inside the mouse embryo, revealing that the middle region is more deformable than the side regions of the inner cell mass.

Moving Path Following (MPF) control laws allows an autonomous vehicle to converge to and follow a path that is moving with respect to an inertial frame. This paper formally extends the MPF methods present in the literature to the case of three dimensional paths that can be moving with time-varying linear and angular velocities with respect to an inertial frame. A 3D MPF error space and a MPF Lyapunov-based control law is derived at kinematic level for a fixed-wing unmanned aerial vehicle. Formal convergence proofs and validation using flight test results are presented, demonstrating the effectiveness of the proposed method.

Cardiac arrhythmias are commonly treated by minimally invasive catheter ablation. Because of challenges in precisely manipulating the distal end of the catheter, several magnetic manipulation systems (MMS) have been developed to aid in steering the tip inside the heart. Due to the relative stiffness of catheters, magnetic gradients that create forces acting on the catheter tip position are limited in their effectiveness and are typically ignored. This dramatically limits the tip orientations possible at any specified position in the workspace. Replacing the flexible catheter tip by a string-like tether and replacing multiple magnets by a single magnet allows the ablation tool to be positioned at any location with essentially any orientation. In addition to the five pose degrees of freedom, the tension on the tether can also be controlled. During an ablation procedure, the distal end of the tethered magnet is placed in contact with the heart wall. Extending the tether after tip contact transfers the tension on the tether to a force at the contact point. The contact force can then be precisely controlled and is no longer dependent on catheter bending radius and insertion length, as is the case with current catheter steering systems. The kinematics of a tethered magnet is modeled and tested inside a clinical MMS called the Aeon Phocus. This demonstrates the feasibility of magnetic field gradient control in a medically certified system currently in clinical use.

We estimate the state of a noisy robot arm and underactuated hand using an implicit Manifold Particle Filter (MPF) informed by contact sensors. As the robot touches the world, its state space collapses to a contact manifold that we represent implicitly using a signed distance field. This allows us to extend the MPF to higher (six or more) dimensional state spaces. Earlier work, which explicitly represents the contact manifold, was only capable of scaling to three dimensions. Through a series of experiments, we show that the implicit MPF converges faster and is more accurate than a conventional particle filter during periods of persistent contact. We present three methods of drawing samples from an implicit contact manifold, and compare them in experiments.

This paper presents a new concept, the Admissible Gap, for collision avoidance. An admissible gap AG is defined as the gap that a robot may safely pass through, while obeying the shape and motion constraints. By employing this concept, a new obstacle avoidance approach was developed, improving the navigation performance in unknown cluttered environments. Unlike most state-of-the-art methods, our approach explicitly considers the robot shape and kinematic constraints rather than adapting a method originally designed for a holonomic pointlike robot. Experimental results demonstrated the power of the proposed AG approach. Moreover, a comparison with state-of-the-art methods showed that the AG approach generates more efficient, safer, and smoother trajectories.

This paper describes the Robotarium - a remotely accessible, multi-robot research facility. The impetus behind the Robotarium is that multi-robot testbeds constitute an integral and essential part of the multi-robot research cycle, yet they are expensive, complex, and time-consuming to develop, operate, and maintain. These resource constraints, in turn, limit access for large groups of researchers and students, which is what the Robotarium is remedying by providing users with remote access to a state-of-the-art multi-robot test facility. This paper details the design and operation of the Robotarium and discusses the considerations one must take when making complex hardware remotely accessible. In particular, safety must be built into the system already at the design phase without overly constraining what coordinated control programs users can upload and execute, which calls for minimally invasive safety routines with provable performance guarantees.

The established processes for measuring physiological and morphological traits (phenotypes) of crops in outdoor test plots are labor intensive and error-prone. Low-cost, reliable, field-based robotic phenotyping will enable geneticists to more easily map genotypes to phenotypes, which in turn will improve crop yields. In this paper, we present a novel robotic ground-based platform capable of autonomously navigating below the canopy of row crops such as sorghum or corn. The robot is also capable of deploying a manipulator to measure plant stalk strength and gathering phenotypic data with a modular array of non-contact sensors. We present data obtained from deployments to Sorghum bicolor test plots at various sites in South Carolina, USA.

To address the goal of locomotion in very complex and difficult terrains, the authors are developing a new class of Ultralight Legged Robots. This paper presents the Multilegged Autonomous eXplorer (MAX), an ultralight, six-legged robot for traversal and exploration of challenging indoor and outdoor environments. The design of MAX emphasizes a low mass/size ratio, high locomotion efficiency, and high payload capability compared to total system mass. MAX is 2.25 m tall at full height and has a mass of approximately 60 kg, which makes it 5 to 20 times lighter than robots of comparable size. MAX is a research vehicle to explore modelling and control of Ultralight Legged Robots subject to flexing, oscillations and swaying; algorithms for gait planning and motion planning under uncertainty; and navigation planning for traversal of complex 3D terrains. This paper presents the design of MAX, provides an overview of the control system developed, summarizes results from indoor and outdoor tests, discusses system performance and outlines the challenges to be addressed next.

A novel wearable robotic device is developed to support the wearer when performing bi-manual tasks near the ground. The device is worn around the upper torso and has two Supernumerary Robotic Limbs (SRLs) that reach the ground when the wearer assumes a crawling-like position, such that neither of the wearer's natural arms are needed to support the body. Rather, both human hands can be engaged in executing a required task. Coordinating actuators at both robotic limbs, the wearer's body is supported stably with a desired impedance. As the wearer moves away from an equilibrium position, restoring forces designated by the impedance act on the wearer. First, the design concept of the SRLs for near-ground work is described, followed by a derivation of the control laws used to produce virtual impedance. A proof-of-concept prototype is then presented along with experimental verification of the derived control laws.

Robotic challenges like the Amazon Picking Challenge (APC) or the DARPA Challenges are an established and important way to drive scientific progress. They make research comparable on a well-defined benchmark with equal test conditions for all participants. However, such challenge events occur only occasionally, are limited to a small number of contestants, and the test conditions are very difficult to replicate after the main event. We present a new physical benchmark challenge for robotic picking: the ACRV Picking Benchmark. Designed to be reproducible, it consists of a set of 42 common objects, a widely available shelf, and exact guidelines for object arrangement using stencils. A well-defined evaluation protocol enables the comparison of complete robotic systems - including perception and manipulation - instead of sub-systems only. Our paper also describes and reports results achieved by an open baseline system based on a Baxter robot.

We present a novel solution to the object search problem for domains in which object permanence cannot be assumed and other agents may move objects between locations without the robot's knowledge. We formalize object search as a failure analysis problem and contribute temporal persistence modeling (TPM), an algorithm for probabilistic prediction of the time that an object is expected to remain at a given location given sparse prior observations. We show that probabilistic exponential distributions augmented with a Gaussian component can accurately represent probable object locations and search suggestions based entirely on sparsely made visual observations. We evaluate our work in two domains, a large scale GPS location data set for person tracking, and multi-object tracking on a mobile robot operating in a small-scale household environment over a 2-week period. TPM performance exceeds four baseline methods across all study conditions.

Two less addressed issues of deep reinforcement learning are (1) lack of generalization capability to new goals, and (2) data inefficiency, i.e., the model requires several (and often costly) episodes of trial and error to converge, which makes it impractical to be applied to real-world scenarios. In this paper, we address these two issues and apply our model to target-driven visual navigation. To address the first issue, we propose an actor-critic model whose policy is a function of the goal as well as the current state, which allows better generalization. To address the second issue, we propose the AI2-THOR framework, which provides an environment with high-quality 3D scenes and a physics engine. Our framework enables agents to take actions and interact with objects. Hence, we can collect a huge number of training samples efficiently. We show that our proposed method (1) converges faster than the state-of-the-art deep reinforcement learning methods, (2) generalizes across targets and scenes, (3) generalizes to a real robot scenario with a small amount of fine-tuning (although the model is trained in simulation), (4) is end-to-end trainable and does not need feature engineering, feature matching between frames or 3D reconstruction of the environment.